I’ve cut task time by 90% in some cases. But I still can’t fully deploy AI in my workflow.

The past 2 months I’ve been experimenting with AI tools and building Agents, but as I take a step back, the same fundamental issues are still there.

It started as a productivity tool. I built a workbook in minutes that would have taken hours.

My mindset shifted. I started using Claude in Excel for simpler tasks and automating workbook rollforward with Claude Cowork.

But it didn’t feel integrated. It felt like I was forcing myself to use this new tool because I knew it was good for me, kinda like remembering at the McDonald’s cash register that the mobile app is faster.

I need the Agent to be integrated with my CRM or spend management tool, ping me or the other stakeholder for review or judgment call, and push the process along in the workflow.

Start with AI and end with AI.

Like when my stomach emits hunger signals, order my McNugget already, and have it delivered. Just let me choose the sauce I’m feeling that day 😂

But today, I’m still exporting data, delivering it to the Agent, generating output, and importing it back.

And that’s just with me and my workflow. Scaling it for the rest of the org is a whole other problem.

Data hygiene. System integration. Team collaboration. Same issues, different tool.

How has your team thought about the integration problem? Not just using AI, but actually deploying it?

"I've already automated a lot of my manual tasks, why invest time, and money, into AI right now?"


Every day I speak to finance professionals at very different points on the AI journey. Some are still figuring out where to begin. Some are using it for everything.

But the group I find most interesting - and probably most underserved by the current conversation - are the leaders who genuinely get it, have already built something good, and are now asking: does AI actually add value when things are already working well?

That question deserves a proper answer. Here's how I'm currently thinking about it.

Using AI for automation gets you from 100 to 50. AI as a tech stack does something different.

In an already-automated finance function, the process is still disjointed. You automate the flag, but someone still reviews, interprets, and acts on it. AI can close that loop - using your business context to not just flag a mismatch, but understand it and commit to an output. Input > insight > action, without the manual stitching in between.

Three areas where I'm seeing this matter most:

1) Workflows - A CFO already running a tight, automated function is well placed to go further. Experimenting with AI is a natural starting point, but the real step-change is when it's genuinely embedded into your workflows rather than bolted on top. That's where you move from "AI saves me time" to "AI runs this process."

2) Finance Systems - A dedicated Finance Systems or FinOps lead who's AI-literate can be transformational here. Not just maintaining your stack, but elevating it - building something that's not just automated but self-auditing and genuinely scalable as the business grows.

3) FP&A and Commercial - A Finance Data Analyst, or equivalent, as the connective tissue between data, finance, and commercial teams. Done well, this isn't about better dashboards - it's about surfacing commercial insight that wouldn't otherwise exist, and adding outsized value at a business level.

Right now (almost definitely) probably isn't the moment to tear up what's working. But it is the training phase, and the teams building these capabilities now - through tooling, through the right hires, or both - will have real options when the technology matures further.

Always interested to hear what people think of this.. feel free to share your thoughts below.

2 years ago, I was scared of AI.

Now I use Claude almost every day in my finance job.

I’m not a tech person. When AI first showed up at work, it felt confusing and overwhelming. And hearing “AI will replace jobs” didn’t help. So I stayed away.

But about 10 months ago, I challenged myself:
👉 What if I just try to learn it?

Not just asking ChatGPT simple questions, but digging deeper and seeing how AI tools can help me save time and do my job better.

At first, there was a lot of trial and error. Some prompts didn’t work. Some outputs didn’t make sense. But slowly, things started to click. Here’s what I’ve learned:

1️⃣ Make the AI prompts as specific as you can.
Being specific minimizes bandwidth and optimizes output.

2️⃣ Think outside the box. Be open to try.
I’ve used it to write, research, extract, analyze, automate, etc.

3️⃣ Understand limitations through testing.
Identify areas of limitation where you need to use judgment.

To me, AI feels a lot like computers decades ago.

Yes, it eliminated jobs. But it also created new ones. And the people who learned how to use computers had the biggest advantage.

AI is the same. Keep an open mind. Try new things. Learn a little every day. Make AI work for you, not against you.

I look forward to continuous learning this year. Curious — how are you using AI in finance today?

I wrote about AI slop back in December.

Marketing emails that sound like robots.
LinkedIn posts that all say nothing.
AI buttons on every app that make everything worse.

It's gotten worse. Way worse.

It jumped departments.

I'm seeing it in spreadsheets now.
Excel models with 20 tabs and yellow input cells on every single one.

Formulas that reference tabs that reference other tabs. Nobody can trace the logic. Nobody checks if it even makes sense.

It looks professional. That's the problem.

Presentations too. Every deck has the same gradient sidebar.

Metrics cards that never end. Total Revenue. Revenue Growth. Revenue Growth Rate. Revenue Growth Rate vs. Prior Quarter. Four cards saying the same thing in different colors. Accent bars across the top like we're designing a magazine.

And a "Key Takeaways" slide that takes away absolutely nothing.

Then the writing. Em dashes everywhere. "The quarterly results were strong — driven by operational efficiency — and positioned the company for growth." That's not a sentence. That's AI doing jazz hands.

Here's what actually happened. People used to be bad at Excel. They knew they were bad at Excel. So they kept it simple. One tab. One purpose. You could follow the logic.

Now AI builds a 20-tab model in 30 seconds and nobody asks if tab 14 needs to exist.

I use AI every day. I run finance, marketing, tech, and HR with it. I build dashboards, automate workflows, draft reports. I'm not some guy yelling at the cloud.

But "AI can build it fast" doesn't mean it should exist.

The whole point was to enhance what we do. Not bring the average down. Not flood every shared drive with pretty garbage nobody reads past tab 3.

More stuff is not better stuff.

Stop it. Think. Then build.

What's the worst AI slop you've seen at work?

Every Claude prompt I use for finance.

30 copy-paste templates. Free.

🙂 Here’s my resource: “30 Copy-Paste Prompts for Claude in Finance”: https://lnkd.in/ghAe4BNj

Most people using Claude for finance?

Getting maybe 20% of what it can do.

Not because the tool is limited.

Because their prompts are garbage.

""Analyze this stock"" → garbage output
""Summarize this filing"" → generic summary

Here's the difference between a bad prompt and a great one:

Bad: ""Analyze Apple's earnings""

Good: ""You are a senior equity research analyst covering mega-cap tech. Analyze the attached Q1 transcript for Apple. Extract: (1) Key financial metrics vs consensus, (2) Management guidance changes with exact quotes, (3) Tone shifts from prior quarter, (4) New strategic initiatives, (5) Top analyst concerns from Q&A. Format as a 1-page brief.""

Night and day.

I compiled my 30 best prompt templates across:

→ Equity Research (5 templates)
→ Investment Banking (5 templates)
→ Private Equity (5 templates)
→ Financial Modeling (5 templates)
→ Claude Code & Automation (5 templates)
→ Pro tips that make EVERYTHING better

Each prompt includes:
✓ The full template (copy-paste ready)
✓ When to use it
✓ Expected output format
✓ Customization tips

The analyst who prompts well ships 5x faster than the one who types ""analyze this.""

You can download “30 Copy-Paste Prompts for Claude in Finance” from: https://lnkd.in/ghAe4BNj

P.S. Pair this with Claude's finance plugins for 10x results.

The scale-up finance teams I'm working with who are pushing the edge of what AI can do are all switching to Claude Cowork. The last months acceleration in this space can't be ignored.

The same AI that writes and runs code now works directly inside your workbooks and across your local files. It reads your models, debugs formulas, cleans messy data, simplifies overcomplex logic, and iterates on its own work.
Inspired by Ruben Hassid's post on building investment bank-grade financial models with Claude in Excel (https://lnkd.in/dptByTBu), I ditched the ChatGPT -> copy-paste -> Excel loop. Now I lean fully into Claude Cowork.

I'm not the only one. Secret CFO described it well - they built a unit economic model feeding an investment case by, in their words, "just streaming semi-coherent thoughts into the prompt box". No formatting. No structuring. Just context in, model out (https://lnkd.in/d2guUdcu). The CFO Office ran a full month-end workflow: vendor contract ingestion, spend-vs-contract cross-referencing, inconsistency flagging, and dashboard creation. Three markdown files, two data exports, zero code. Fifteen minutes to set up.

Here's what I tested:

I used (the ring fenced enterprise version of it that keeps your data private) Cowork on our Actuals, our context, our terminology. After 10 minutes of prompting and some Q&A it asked for, it built me a reforecast and valuation model in 20 minutes.
One that actually made sense.

That same work used to take me (optimistically) roughly a week in calendar time once you factor in meeting blocks, context switching, and iteration cycles, and 2-3 FTEs.
This is the first time I've seen AI efficiently integrate into the native workflow for FP&A and Corporate Finance. Not a chatbot you paste into for inspiration. Not a copilot that understands your spreadsheet maths but not your business logic. An agent that works across your files, understands your context, and produces real outputs.

On the Controllership/Accounting side, we've been fortunate to work with the team at Stacks for deployed AI. Now, impactful solutions are finally reaching the rest of the finance stack as well, Anthropic just shipped five open-source finance plugins (financial analysis, IB, equity research, PE, wealth management) with connectors to FactSet, S&P Capital IQ, and LSEG built in.

If you're still copy-pasting between a chatbot and Excel, or running this all yourself as we did for the last decade, try this. Anything you do monthly in Excel, and in operational finance that's a lot of shuffling, can now be automated in your native workspace. That unlocks time for analysis, controlling, and the value-adding work that actually matters.
Looking forward to hooking up and giving the Netsuite connector to Claude a spin this week!

The workflow shift is material.

I've been using AI in my finance workflows for a while now. But about 60 days ago something shifted. The output stopped looking like AI output. The analysis was sharp, the structure was clean, and for the first time I felt comfortable putting my name on the work and sending it to senior leadership and key stakeholders. Anyone in finance knows what that bar feels like.

What used to take hours to sometimes days now takes minutes. But speed isn't the real story. The real story is depth. I'm running deep dives on problems I never had the bandwidth to go past surface-level on. Building scenario models with dozens of variables and layered sensitivities. Fully automating end-to-end reporting processes. Getting real-time analytical feedback on board narratives before they ever reach the boardroom.

But the tool isn't why it's working. The reason it's working is that I'm in it. Every day. Learning, building, breaking things and rebuilding them. Not hiring a vendor to spend six figures, build something nobody understands, and leave behind a tool nobody uses.

AI adoption in finance only works when the finance person IS the AI person. Domain expertise and tool expertise have to sit in the same chair. You can't outsource that. It's a skillset, not a software implementation.

Anthropic recently released Cowork, which lets AI work across your files and workflows autonomously. I've been pressure testing it daily, and this week I'm pushing it hard to see how far it can expand into other areas of the finance function. It feels like an arms race to stay ahead of the curve right now, and the gap between finance leaders who are learning this and those who aren't is widening by the week.

I'm a CPA with 12+ years across EY, Fortune 100 FP&A, and PE-backed operations. I've sat through a lot of vendor pitches. This is the first time the competitive advantage goes to the practitioner who learns it, not the company that sells it.

If you're in strategic finance and want to compare notes, let's connect. Getting started is the hardest part, and I'm happy to share what I've learned so far.

The tools are ready. The question is whether finance leaders are willing to get their hands dirty.